Causal Triplet

该数据集是一个因果表征学习基准，它特别突出了视觉上复杂的场景以及具有实际操作性的反事实设定。该数据集包含了具有组合性变化和系统性变化的场景设置，重点关注单一对象和多对象场景。其任务旨在研究和建模因果表征学习及干预。

This dataset is a causal representation learning benchmark that specifically highlights visually complex scenes and practically actionable counterfactual settings. It features scenario setups with both compositional variations and systematic variations, with a primary focus on single-object and multi-object scenes. The tasks included in this benchmark are designed to investigate and model causal representation learning and interventions.